The Research Of Red Tide Algae Images Classification And Recognition Technology

The Red Tide is one of the serious familiar disasters along the offing, which not only causes great loss financially, but holds severe threat against oceanic environment and its resources as well as public health. While the loss led by the Red Tide could only be lessened through inspecting and forecast at present. Therefore, it is urgent for us to have an effective inspecting method for the Red Tide Algae. The traditional inspecting method is to employ microscopes to make identification and take count, by which there are not only intense labor and inefficiency, but also it is hard to differentiate due to the close forms of the Red Tide. So the task of classification and identification could be accomplished by experienced experts only. All of these factors thus would have severe impact upon the responsive time to the forecast of the Red Tide, quite unfavorable to the lessening and preventing of the disaster of the Red Tide.This paper mainly studies to bring forward a method of a veracious, real-time automatic classification for the Red Tide Algae. Firstly it makes a feature analysis of the original data of Red Tide Algae Images and receives the optimal feature subset by a feature selection among the feature set with the intension to remove those unrelated and redundant features. Different classifying effects of SVM and KNN on the optimal feature subset are discussed and analyzed respectively afterwards. Finally the paper claims to use the SVM-KNN classifier to make the identification and classification of Red Tide Algae Images.The main points as well as the creative ones in the paper are included in the following:(1) Based on the data analysis of the original feature set, the paper brings forward the feature selection method by combining the ReliefF algorithm and sequential backward search strategy, which can remove the unrelated and redundant features in the original feature set and lessen the impact upon the precision of classifiers. Besides, the experiment data show the comparison of the classifying effects both before and after classifying by SVM and KNN.(2) SVM and KNN are employed respectively to conduct classificatory experiments on four types of sample data set and seven types of sample data set received respectively after feature selection. Features and properties of the two classifiers are later discussed and analyzed according to the results of the experiments.(3) The paper conducts study of the theories of SVM and KNN classifiers and its theoretically common points have been discovered. Hence the theoretical foundation of combining both classifiers is obtained and so the paper claims to adopt the SVM-KNN classifier to improve the classifying effect on the red tide algae image. Furthermore, experiment data prove the truth that the SVM-KNN classifier can enhance the classifying performance effectively.